Skip to content

cp: fix: AC silently skipped on all registered VLMs β€” flatten ModuleList (1941) into r0.4.0#1958

Merged
akoumpa merged 1 commit intor0.4.0from
cherry-pick-1941-r0.4.0
Apr 21, 2026
Merged

cp: fix: AC silently skipped on all registered VLMs β€” flatten ModuleList (1941) into r0.4.0#1958
akoumpa merged 1 commit intor0.4.0from
cherry-pick-1941-r0.4.0

Conversation

@svcnvidia-nemo-ci
Copy link
Copy Markdown
Contributor

beep boop [πŸ€–]: Hi @khazic πŸ‘‹,

we've cherry picked #1941 into  for you! πŸš€

Please review and approve this cherry pick by your convenience!

…1941)

fix: flatten ModuleList in _extract_model_layers so AC applies to individual layers

_reduce_attrs returns ModuleList objects as single items; extending layers
with them meant AC code never found self_attn/mlp on a ModuleList and
silently skipped all checkpointing. Flatten any ModuleList results so
layers contains individual decoder layers, matching the heuristic path.

Signed-off-by: khazic <khazzz1c@gmail.com>
Signed-off-by: NeMo Bot <nemo-bot@nvidia.com>
@svcnvidia-nemo-ci
Copy link
Copy Markdown
Contributor Author

/ok to test 8566f86

@copy-pr-bot
Copy link
Copy Markdown

copy-pr-bot Bot commented Apr 21, 2026

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@akoumpa akoumpa merged commit f23c168 into r0.4.0 Apr 21, 2026
51 of 54 checks passed
@akoumpa akoumpa deleted the cherry-pick-1941-r0.4.0 branch April 21, 2026 19:50
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

cherry-pick Run CICD Trigger Testing CICD

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants